refactor: worker mode — server offloads all LLM/exec to worker

- Split into `tori server` / `tori worker` subcommands (clap derive)
- Extract lib.rs for shared crate (agent, llm, exec, state, etc.)
- Introduce AgentUpdate channel to decouple agent loop from DB/broadcast
- New sink.rs: AgentUpdate enum + ServiceManager + handle_agent_updates
- New worker_runner.rs: connects to server WS, runs full agent loop
- Expand worker protocol: ServerToWorker (workflow_assign, comment)
  and WorkerToServer (register, result, update)
- Remove LLM from title generation (heuristic) and template selection
  (must be explicit)
- Remove KB tools (kb_search, kb_read) and remote worker tools
  (list_workers, execute_on_worker) from agent loop
- run_agent_loop/run_step_loop now take mpsc::Sender<AgentUpdate>
  instead of direct DB pool + broadcast sender

src/template.rs

@@ -3,7 +3,6 @@ use std::path::{Path, PathBuf};
 use serde::Deserialize;
 
 use crate::TemplateRepoConfig;
-use crate::llm::{ChatMessage, LlmClient};
 use crate::tools::ExternalToolManager;
 
 #[derive(Debug, Deserialize)]
@@ -463,42 +462,6 @@ pub fn is_repo_template(template_id: &str) -> bool {
     template_id.contains('/')
 }
 
-// --- LLM template selection ---
-
-pub async fn select_template(llm: &LlmClient, requirement: &str, repo_cfg: Option<&TemplateRepoConfig>) -> Option<String> {
-    let all = list_all_templates(repo_cfg).await;
-    if all.is_empty() {
-        return None;
-    }
-
-    let listing: String = all
-        .iter()
-        .map(|t| format!("- id: {}\n  名称: {}\n  描述: {}", t.id, t.name, t.description))
-        .collect::<Vec<_>>()
-        .join("\n");
-
-    let prompt = format!(
-        "以下是可用的项目模板：\n{}\n\n用户需求：{}\n\n选择最匹配的模板 ID，如果都不合适则回复 none。只回复模板 ID 或 none，不要其他内容。",
-        listing, requirement
-    );
-
-    let response = llm
-        .chat(vec![
-            ChatMessage::system("你是一个模板选择助手。根据用户需求选择最合适的项目模板。只回复模板 ID 或 none。"),
-            ChatMessage::user(&prompt),
-        ])
-        .await
-        .ok()?;
-
-    let answer = response.trim().to_lowercase();
-    tracing::info!("Template selection LLM response: '{}' (available: {:?})",
-        answer, all.iter().map(|t| t.id.as_str()).collect::<Vec<_>>());
-    if answer == "none" {
-        return None;
-    }
-
-    all.iter().find(|t| t.id == answer).map(|t| t.id.clone())
-}
 
 // --- Template loading ---